Clinical Research Oriented Workshop (CROW) Meeting: Jan 17, 2013

Present: Abby Crocker, Kairn Kelley (by phone), Amanda Kennedy, Rodger Kessler, Ben Littenberg, Charlie MacLean, Prema Menon, Connie van Eeghen

1.                  Start Up:  News from Rodger: instructions from study section leader – all applications start with a score of 5 (middle of the range). This is the result of clustering at the low (better) end. 

2.                  Presentation: Future CROW group project

a.       In a recent CROW session, the group had agreed that it would be interesting to discuss a broad outline of a group research project with roles organized around investigator, analyst, and reviewer.  One of our goals is to learn more about multi-level modeling through a team-based research study. The ideal project will result in a publication.

b.      Possible topics:

                                                  i.      Benzos: currently topical due to an FDA announcement on Monday restricting prescription dosages for women.  High prevalence; alternative treatments are present in data bases.

                                                ii.      Stimulants for children with ADHD (short-acting vs. long-acting)

c.       Data sources:

                                                  i.      Data are available now from VHCURES without Medicare age group; may be missing many clinical co-variates (limited to what is paid for by insurance).  Submit a request to Liz and work with Liz and Steve Kappel to organize.  IRB moves quickly and Steve runs the data. Almost five years of data (started in 2007).

1.      Is the denominator patients or prescribers?  Patients will have incomplete histories based on cash payments and out-of-state services. There may be problems with unique patient identifiers, as insurance companies have different patient codes.

2.      Topics must be exclusive of geriatrics and hospital admissions.  Medications and procedures only. Does have ED visits.  Does not have zipcodes.

                                                ii.      IRIS: many more co-variates (entire clinical record); no claims from VHCURES with more extensive med list.  PRISM med list is smaller and less accurate; patients may be getting care that is invisible to IRIS.

d.      Research topic: “what is the association of x to y while controlling for z.”  We’re interested in topics that are not just driven by the main predictor but cluster by certain variables (e.g. providers).  Therefore, a larger sample size isn’t as important as the number of providers, and the sample sizes within those clusters.  This phenomenon can appear at many levels: practice, county, and within patient (for repeated events, such as cardiac care visits).  Consider picking a homogenous population to provide a good template for this type of research study and a predictor with greatest opportunity to learn about different analytical models.

                                                  i.      Motivation for Kairn, who started us down this road: each child is tested more than once, from a specific class, on a specific day of week.  Different analytical models produce different results; when to use which?

                                                ii.      Possible predictors: geography, type of practice (primary, specialty), medication (hypnotic meds [benzos] dose category, short acting opiates, babies on opiates, multi-prescriber opiates, stimulants long/short)

1.      The team is considering opiates, categorized by long/short, length of time (< and > 60 days).  Could have a systematic intervention (duration limits on short-acting Rx)

                                              iii.      Possible outcomes: ED visits, use of medications (atypicals, benzos, opiates), event related to medication, surgeries, diagnosis codes related to procedures, cost/utilization in a defined time period, fall/fractures, dose progression, persistence of drug therapy (point prevalence), drug substitution, clinical momentum (continued use of drugs on auto-pilot)

1.      The team is considering ED visits

                                              iv.      Variables to control: time (secular trends in prescribing)

                                                v.      Lines of inquiry:

1.      How much of the variation of x is explained by provider vs. patient.

2.      Comparative effectiveness

3.      Penetration of interventions

4.      Dose/response effect: benzos and ED visits

a.       Classify into high/med/low dose

b.      Adults only (age restriction)

c.       Analyze event rates for three dosage categories

5.      Natural history of drug use

                                              vi.      Next step: VHCURES code book; Steve Kappel – Connie to follow up

3.                  Next Workshop Meeting(s): Thursday, 2:00 p.m. – 3:30 p.m., at Given Courtyard Level 4. 

a.       Jan 24: Abby: breast feeding paper

b.      Jan 31: Kairn: F31 update

c.       Feb 7: Abby:

d.      Feb 14:

e.       Future agenda to consider:

                                                  i.      Christina Cruz, 3^rd year FM resident with questionnaire for mild serotonin withdrawal syndrome on 12/6 or 12/13

                                                ii.      Peter Callas or other faculty on multi-level modeling

Recorder: Connie van Eeghen